Medical AI may be stuck in a bind familiar to many jobseekers: If you don’t have the hands-on experience, you can’t get the job. If you don’t have the job, you can’t get the hands-on experience.
Framed another way, the concept of medical AI has attained widespread acceptance almost as fast as the successful studies have proven it out. And the nods have come from the general public as well as the clinically sophisticated.
However, AI is still not ready for daily practice because the technology’s success record has been built mostly in research settings.
That’s a paraphrase of a literature review published online Aug. 3 in the German journal Der Hautarzt (The Dermatologist) and lead-authored by Titus Brinker, MD, of the German Cancer Research Center in Heidelberg.
According to the study abstract as rendered in English by Google Translate, Brinker and colleagues extrapolated the broad healthcare landscape from studies testing AI diagnosis of skin cancer.
The team searched the literature for recent trials relevant to that concentration using keywords such as artificial intelligence, machine learning and, with particular keenness, convolutional neural networks, which “have been shown to be particularly effective for the classification of image data,” the authors note.
Analyzing the results, Brinker et. al found that, in multiple studies, AI algorithms have detected and distinguished possible skin cancers “with high precision, comparable to that of dermatologists.”
The best results have come when AI tools have worked in tandem with physicians’ examinations of their patients.
Nevertheless, because of the lack of utilization beyond the academy, “many digital diagnostic criteria that help AI to classify skin lesions remain unclear,” the authors comment. “This lack of transparency still needs to be addressed. Furthermore, clinical studies on the use of AI-based assistance systems are needed in order to prove its applicability in daily dermatologic practice.”
And by extension in daily practice across the spectrum of medical specialties.